Hello. I think there was a bug in my AI Pro subscription today:
Today, when my daily usage quota was completely reset, I started a new conversation with a completely new quota and well:
After about several low requests, I hit the daily usage quota with about 5-6 requests. That was with an active AI Pro subscription. It was exactly like the free trial quota and I reached the limit quickly. However, I had AI Pro and I always reached this daily limit after several hours and many and heavy requests.
I’m also getting the same issue, what is the flag you are getting with your response?
I’m on the first Ultra tier with Ai Sub, but I notice its stating free tier in the flag. Mostly using 3.8 flash, 5-6 requests as well before hitting rate limits.
‘Failed to generate content: user has exceeded quota. Please try again later.’
‘warning You are out of free generations.
You can continue by using your own paid API key.’
Same here. This started happening for me right at the beginning of September (I think during the new flash model’s release). I’m on the AI Pro plan and I’m getting rate-limited on AI Studio after just 4 prompts today, which is insane considering last week I was easily running 60+ prompts a day through heavy context chats without a single hitch.
Now the platform clamps shut instantly, even on tiny 2k windows with short prompts. There’s zero heavy usage or big files on my end right now, yet it still forces a full 24-hour lockout after a couple of prompts. No spare messages between the hours either unlike before, just forces you to wait out the day. Getting hard-capped immediately like this is completely ridiculous for a paid sub. I’m not sure why more people aren’t reporting this issue.
I suspected exactly the same thing. Despite an active AI Pro subscription, when I hit a quick limit after 5 or 6 requests, like the Free Trial limit, I got a few errors. One was that your daily quota has been exhausted, and another that was very suspicious was that it said: You are out of free generations. You can continue by using your own paid API key.
I can’t believe it wasn’t enough yesterday, and today the same thing happened with a completely reset quota. Even though I still have my subscription and it’s active, it’s giving me exactly the errors you see in the picture. It’s really ridiculous.
I’m unsure if I’m remembering this correctly or not, but those specific pop-up messages; the ‘You are out of free generations’ banner is just a generic, hard-coded UI alert in AI Studio. It pops up for absolutely everyone whenever a daily model cap or token limit is reached, regardless of whether you’re on a paid tier or free.
The wording is a bit misleading because it makes it look like a free-tier error, but it’s just the default fallback screen when the standard query bucket empties for the day. The core issue we’re seeing is simply the tight limits on the models themselves.
To back that up, a lot of other users (from different sites as well) are reporting the exact same sudden drop in caps over the last 48 hours. It looks like this throttling kicked into overdrive right as the new 3.8 models dropped. Though there’s not much else I can pick up on it rather than my own cases and a few others.
I agree, and yes, it’s true that you can’t tell from a single popup message alone. However, what we do have in common is that the limits have increased dramatically over the past 48 hours and recently in general.
As you mentioned, we previously had more than 50 heavy requests with high token usage, but recently that has dropped to just 5 or 6 short requests, almost as if we don’t have a subscription at all, despite having an active AI Pro subscription.
Your point, and the point others have made, that this started happening around the release of Flash 3.8 is interesting. However, nothing is confirmed at this point, and the Google team needs to address this and provide an explanation.
Completely agree. What concerns me right now is the silence on this, with no official word or status update yet, it feels like it’s flying completely under the radar. Since most general users are on the mainstream Gemini platforms (which use short-term rolling caps), a specific infrastructure issue like this in AI Studio just isn’t getting the noise it needs to be acknowledged. I really hope the moderation or engineering team spots these threads soon. It’s hard not to worry that a critical backend glitch affecting paid subscribers might just sit there unnoticed unless we get some official eyes on it, as I’d rather not see this become the new baseline and remain disruptive to people’s established workflows ;
The interesting thing that I just realized and I think you need to know is that my AI Pro subscription ended just a few hours ago, and now I renewed it again for 1 month. And well, it was expected that there would be no rate limit, after all, the new subscription was just activated. But what was expected did not happen, and a suspicious thing happened: I am still completely rate-limited. This makes me more suspicious of and closer to the hypothesis that I am being given the free-tier quota while the newly renewed AI Pro subscription is active.
Now it really feels like my money is being wasted, the money I paid for the subscription, and it makes me regret it. And well, this issue is really serious and should be investigated immediately by the Google team.
On a semi-related note, even if the message itself is worded as if the user is on a free tier… if 3.8 Flash is so much cheaper than 3.1 Pro, why does 3.1 Pro have a much, much more generous quota? It’s Google’s business after all, but 4 prompts in with 3.8 Flash and I’m out. What’s surprising is that this is exclusive to the Build mode. It’s not something that’s present in the Playground as far as I can tell, but I haven’t tested it that extensively to confirm for 100%. (edit: I am on Google AI Pro)
I heard that 3.8 Flash uses far more tokens… while staying cheaper overall. It doesn’t seem super fair to me to lock users out like that, but again, this is not the place to complain about Google’s business model.
That’s strange, my case happens to be different. I was actually running 3.1 Pro and getting hit with the exact same 4 (or something around that) prompt wall. Switched between the flash models over the days. It seems completely indiscriminate across all models/or not on a model by model basis.
I can’t speak for the Build Mode vs. Playground distinction, but just wanted to throw my data in.
Sorry to crowd this thread with my rant but.. As a paid Pro tier subscriber, I’ve had the same issues too.
I’m a relatively new user to this platform, since when I originally sought help in places like the AI Studio or Bard subreddit. The responses there were frustratingly unhelpful, I was essentially gaslit and told the problem was on my end, that I needed to cut down my context windows, switch models, or that the issue itself wasn’t even real. But Like others have said before September, Even during peak demand periods, the platform stayed completely stable and responsive. Yet now nobody even wanted to point out the original problem or differences.
Now??? I can’t even get past over 8 prompts or less or get any actual work done. To prove this isn’t user error, I literally tested the exact same prompts in the exact chats on a separate Free-tier Account and discovered there is absolutely zero difference in behavior or limitations right now. Paying for a Pro subscription is currently yielding the exact same restricted utility as a free tier user like everyone else has theorized here.
I’ve completely lost all hope and have no idea what I’m even paying for anymore. I may fully unsub this week if they intentionally plan to keep this around because there is no difference between the free tier or the planned subs.
Exactly like you, I use 3.1 Pro more, and it doesn’t matter which model. Both 3.1 Pro and 3 Flash reached the daily limit after about 5 or 6 requests with an active paid subscription, so it’s kind of the same for all models. I agree.
It’s very interesting. This is exactly the same thing I felt when I ran into this problem and decided to create this Topic. I literally felt like I was on the free plan, both when my subscription was nearing its end and when I renewed my subscription. And well, with your test, there is no doubt left anymore. The rights of those who paid for a subscription are being violated.
Another day with this problem… exactly after 8 requests, with another day of the Pro subscription being wasted, not only for me but for all of us, and another day of projects being delayed… and with no answer from the Google team… Every day that this problem continues to exist, I will repeat the status in this Topic every day. I will repeat it every day and keep the Topic alive until we get an answer…
Please do. I’ll try to be bumping this thread daily to log my run-outs too until someone from the team addresses it.
This situation is really unprecedented because looking back at it now, When Google launched the recent flash models back in the summer, paid tiers didn’t suddenly have 95% of their capacity capped overnight. (Making a paid subscription essentially useless is a bizarre business move, Even the pro releases weren’t this restrictive.) If this is the long-term plan, the team should atleast give us a straight answer. Users deserve full transparency on whether these changes are permanent so we can decide what to do with our subscriptions.
I have an active Google AI Pro subscription and I use AI Studio for long-form creative writing, mostly with 3.1 Pro and large context. Until a few days ago, a single session would run roughly 5-6 hours of continuous work before I hit the cap. Since around September 5-6, the same workflow gets me about 1,5 hours, sometimes less, and then I’m locked out completely.
Nothing changed on my side: same model, same project, same kind of prompts, same subscription. The drop was sudden, not gradual.
Screenshot attached. Would really appreciate some official word on whether this is a backend issue with how paid subscriptions are being classified, or an intentional change to the limits.
Not to go off topic but technically… Google can do whatever they want with these quotas. Their AI Studio docs say capacity isn’t guaranteed and that limits can change without warning. They’ve done it before. I’m not defending its practice, but it is what it is. Tho It goes back to what you said, it’s one thing to cut some free users off to save server space during a model rollout. It’s an entirely different level to clamp down on paid subs after just a few prompts without a single word of explanation. Right now, the platform isn’t even functional for paying customers.
